Utilization review in nursing homes: making implicit level-of-care judgments explicit

Area of Science:

• Gerontology
• Health Services Research
• Biostatistics

Background:

• Nursing home patient care level (LOC) assessment is crucial.
• Accurate LOC determination impacts resource allocation and patient outcomes.
• Current methods may be resource-intensive and subject to bias.

Purpose of the Study:

• To compare algorithmic and logistic regression models for predicting appropriate patient LOC in nursing homes.
• To assess the accuracy and utility of these models for preliminary patient screening.

Main Methods:

• Trained observers collected data on 3,579 nursing home patients.
• Two approaches were tested: an algorithm based on clinical criteria and logistic regression equations.
• Logistic regression models were trained on half the data and tested on the other half.

Main Results:

• Both approaches yielded comparable results.
• The best algorithm correctly identified 71% of skilled care and 69% of unskilled care patients.
• Logistic regression models correctly classified 86% of skilled care and 63% of non-skilled care patients.
• Optimizing prediction for one group sometimes decreased accuracy for the other.

Conclusions:

• Simple models, particularly logistic regression, show promise for preliminary nursing home patient LOC screening.
• These methods can reduce the demand for skilled professional judgment.
• Recommended for preliminary screening, not final LOC determination.